INDEPENDENT VERIFICATION

Is it really being checked?
See it for yourself.

Trust shouldn't rest on the developer's word. So we run two tests and publish the results exactly as they came out.

INDEPENDENTLY VERIFIED · PASS

Independent verification report (v0.9.0)

5,005files verified

0mismatches

4tools cross-checking

TEST 1

A normal copy — does every file match?

We compared the “fingerprint” of all 5,005 files between the source and the destination.

5,005files → ✓ All match (0 mismatches)

TEST 2

Break files on purpose — does it notice?

We deliberately corrupted 10 files and copied them, to see if the checks would miss any.

broke file 1 ✓ caught broke file 2 ✓ caught broke file 3 ✓ caught broke … ✓ broke file 10 ✓ caught

10all of them → ✓ Caught (0 missed · 0 false positives)

We broke them in different ways too: flip a byte / truncate / append. It still catches every one.

THE CHECKERS

Real, unrelated tools do the checking

The app saying “OK” on its own is just self-grading. So four tools with no connection to MasterCopy grade the copy independently — each a real inspection tool used by professionals worldwide.

hashdeepA standard file-auditing tool — used even in digital forensics

rsyncThe standard sync tool on Mac / Linux (compares contents to find differences)

xxhsumThe official tool for the fast xxHash

ascmhlThe official tool for the film-industry standard ASC MHL

AND

Not one check, but several — each confirms separately

The same 10 corrupted files, each detected independently.

Corrupted file	MasterCopythe app	hashdeep3rd-party audit	xxhsumfast hash	SHA-256standard hash
zero-byte.bin	✓	✓	✓	✓
日本語素材／カメラ A・本番テイク01.mov	✓	✓	✓	✓
DCIM/A001/A002_C4251_004251.braw	✓	✓	✓	✓
Photos/IMG_1770.JPG	✓	✓	✓	✓
… 6 more	✓	✓	✓	✓

✓ = detected / all 10 · every tool · 0 false positives

THE TEST DATA

We tested with deliberately “nasty,” real-world-like data

It's synthetic data, but we packed in the conditions that actually trip things up on a real shoot.

Japanese & multilingual filenames (full-width characters, spaces, middle dots, symbols)
Camera card structures and real formats: DCIM / .braw / .mov / .wav / .jpg
Nasty names: symbols #&() []/clip #1 & take (final).wav
Edge cases: very deep folders, very long names, zero-byte files

The bytes are reproducible synthetic data; the names, structure, formats and sizes mirror real material. If Japanese and symbols pass, filenames in any language work the same way.

THE REAL THING

The actual report, auto-generated by GitHub

This is the real report behind the summary above. The scripts, run logs and results are all published on GitHub. If you're curious, you can check it for yourself.

Open the full reportThe report generated by GitHub Actions GitHub Actions run logsRun logs on public CI (hosted by GitHub) Raw logs (4-tool output)Raw output from xxhsum / hashdeep / rsync / ascmhl All scripts (repository)Verification scripts and report.json (source of truth)

Everything above runs on GitHub Actions (a throwaway inspection server); the scripts, logs and results are published on GitHub. The run log is hosted by GitHub, so the author can't rewrite it after the fact. The tests use synthetic data generated deterministically from a seed, so anyone can reproduce the exact same result. The bytes are synthetic, but the names, structure, formats and sizes mirror real material — and when you handle real footage, the checking works the same way: a byte-by-byte comparison of every file.

Is it really being checked?See it for yourself.